THEORY EXAMINATION (SEM–VI) 2016-17 INFORMATION RETRIEVAL AND MANAGEMENT

INFORMATION RETRIEVAL AND MANAGEMENT (NIT061)

SECTION – A

(Short Answer Questions | 10 × 2 = 20 Marks)

(a) Rocchio’s Classification Method

Rocchio’s method is a relevance-feedback based classification technique. It constructs a prototype vector for each class by averaging document vectors belonging to that class. A new document is classified into the class whose prototype vector is closest (usually using cosine similarity).

(b) Uses of Dendrograms & Hierarchical Clustering

Dendrograms visually represent hierarchical clustering. They show how documents or data points are grouped step-by-step based on similarity. They are used to analyze cluster structure and decide the optimal number of clusters.

(c) Precision–Recall Curves

Precision–Recall curves show the trade-off between precision (relevant retrieved / total retrieved) and recall (relevant retrieved / total relevant). They are used to evaluate and compare IR system performance.

(d) OPAC vs Information Retrieval Systems

OPAC (Online Public Access Catalog) is mainly used in libraries for searching books using metadata like author or title.
IR systems retrieve unstructured information (documents, web pages) using content-based search.

(e) SIMD vs MIMD

SIMD executes the same instruction on multiple data streams simultaneously.
MIMD executes different instructions on different data streams, making it more flexible and suitable for parallel IR systems.

(f) Performance Measures of Parallel Computing

Key performance measures include speedup, efficiency, scalability, throughput, and response time.

(g) Parallel Multitasking on MIMD

In MIMD systems, multiple processors execute different tasks independently. Tasks are distributed dynamically, allowing parallel query processing and indexing in IR systems.

(h) Query Routing Problem

Query routing determines which databases or nodes in a distributed IR system should receive a query. Efficient routing improves response time and retrieval effectiveness.

(i) Local vs Global Inverted File

Local inverted file stores index information separately at each node.
Global inverted file maintains a single index for all documents across nodes, simplifying query processing but increasing communication cost.

(j) Design Issues of Crawl Module

Key issues include URL selection, politeness policy, duplicate detection, scalability, freshness of data, and fault tolerance.

SECTION – B

(Attempt Any Five | 5 × 10 = 50 Marks)

(a) Challenges in Information Retrieval Systems

IR systems face challenges such as information overload, vocabulary mismatch, ambiguity, relevance judgment, scalability, and user behavior diversity.
These can be addressed using query expansion, relevance feedback, machine learning, semantic search, and efficient indexing techniques.

(b) Signature File Approach

The signature approach represents documents using bit patterns (signatures). It is faster and requires less storage compared to full inverted indexes.
Types include word-signature, document-signature, and superimposed coding based methods.

(c) Challenges in Digital Library Architecture

Challenges include data heterogeneity, scalability, metadata management, copyright issues, preservation, user interface design, and efficient retrieval mechanisms.

(d) MIMD Architecture in Parallel IR

In MIMD architecture, processors operate independently with their own memory or shared memory. Queries and indexing tasks are distributed among processors, improving speed and fault tolerance in large IR systems.

(e) Working of a Search Engine

A search engine works in three main stages:

Crawling – collecting web pages

Indexing – organizing data using inverted indexes

Query processing – matching user queries with indexed data and ranking results

(f) Multimedia Information Retrieval

Multimedia IR retrieves images, audio, and video using content features like color, texture, shape, and metadata. Retrieval relies on feature extraction, indexing, similarity matching, and relevance ranking.

(g) Query Processing in Distributed IR

Query processing involves query decomposition, routing, local retrieval at nodes, merging results, and ranking. Efficient coordination reduces communication overhead.

(h) Historical Development of Information Systems

Information systems evolved from manual record keeping to computerized databases, management information systems, decision support systems, and modern web-based and intelligent IR systems with AI integration.

SECTION – C

(Attempt Any Two | 2 × 15 = 30 Marks)

3. OPACs and Their Limitations

OPACs provide online access to library catalogs, allowing users to search books, journals, and media.
Limitations:
They rely on metadata rather than content, offer limited search flexibility, lack relevance ranking, and are not suitable for unstructured information retrieval.

4. Diet Expert System – Case Study

A diet expert system stores data such as:

User profile (age, weight, health condition)

Nutritional values of food items

Dietary rules and constraints

Efficient retrieval methods include rule-based inference, indexed databases, decision trees, and query optimization techniques to provide personalized diet recommendations.

5. MIMD Architecture with Explanation

MIMD architecture consists of multiple processors executing different instructions on different data.
Each processor may have its own memory or share memory.
It supports multitasking, parallel query processing, and large-scale IR applications due to flexibility and scalability.